Skip to content

whisper : reduce delta_min from 1000ms to 100ms #3028

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Apr 11, 2025
Merged

whisper : reduce delta_min from 1000ms to 100ms #3028

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 11 additions & 9 deletions src/whisper.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5527,11 +5527,13 @@ int whisper_full_with_state(
const int seek_start = params.offset_ms/10;
const int seek_end = params.duration_ms == 0 ? whisper_n_len_from_state(state) : seek_start + params.duration_ms/10;

// if length of spectrogram is less than 1.0s (100 frames), then return
// basically don't process anything that is less than 1.0s
// see issue #39: https://github.com/ggml-org/whisper.cpp/issues/39
if (seek_end < seek_start + 100) {
WHISPER_LOG_WARN("%s: input is too short - %d ms < 1000 ms. consider padding the input audio with silence\n", __func__, (seek_end - seek_start)*10);
// if length of spectrogram is less than 100ms (10 frames), then return
// basically don't process anything that is less than 100ms
// ref: https://github.com/ggml-org/whisper.cpp/issues/2065
const int delta_min = 10;

if (seek_end < seek_start + delta_min) {
WHISPER_LOG_WARN("%s: input is too short - %d ms < 100 ms. consider padding the input audio with silence\n", __func__, (seek_end - seek_start)*10);
return 0;
}

Expand Down Expand Up @@ -5675,8 +5677,8 @@ int whisper_full_with_state(
ctx, state, progress_cur, params.progress_callback_user_data);
}

// if only 1 second left, then stop
if (seek + 100 >= seek_end) {
// if only 100ms left, then stop
if (seek + delta_min >= seek_end) {
break;
}

Expand Down Expand Up @@ -6023,10 +6025,10 @@ int whisper_full_with_state(
// end of segment
if (token.id == whisper_token_eot(ctx) || // end of text token
(params.max_tokens > 0 && i >= params.max_tokens) || // max tokens per segment reached
(has_ts && seek + seek_delta + 100 >= seek_end) // end of audio reached
(has_ts && seek + seek_delta + delta_min >= seek_end) // end of audio reached (100ms)
) {
if (result_len == 0 && !params.no_timestamps) {
if (seek + seek_delta + 100 >= seek_end) {
if (seek + seek_delta + delta_min >= seek_end) {
result_len = i + 1;
} else {
WHISPER_LOG_DEBUG("%s: decoder %d failed (result_len = 0)\n", __func__, j);
Expand Down
Loading